Distributed localization of networked cameras

Abstract

Camera networks are perhaps the most common type of sensor network and are deployed in a variety of real-world applications including surveillance, intelligent environments and scientific remote monitoring. A key problem in deploying a network of cameras is calibration, i.e., determining the location and orientation of each sensor so that observations in an image can be mapped to locations in the real world. This paper proposes a fully distributed approach for camera network calibration. The cameras collaborate to track an object that moves through the environment and reason probabilistically about which camera poses are consistent with the observed images. This reasoning employs sophisticated techniques for handling the difficult nonlinearities imposed by projective transformations, as well as the dense correlations that arise between distant cameras. Our method requires minimal overlap of the cameras' fields of view and makes very few assumptions about the motion of the object. In contrast to existing approaches, which are centralized, our distributed algorithm scales easily to very large camera networks. We evaluate the system on a real camera network with 25 nodes as well as simulated camera networks of up to 50 cameras and demonstrate that our approach performs well even when communication is lossy.

Distributed Maximum Likelihood for Simultaneous Self-Localization and Tracking in Sensor Networks

“…θ r,j . The local joint predicted density p r θ (x r n |Y 1:n−1 ) at node r was defined in (8) and is a function of θ = {θ i,j } (i,j)∈E , and likelihood term is given in (9). Also, the gradient is evaluated at θ n = {θ i,j n } (i,j)∈E while only θ r,j n is available locally at node r. The remaining values θ n are stored across the network.…”

“…However, our work differs from [9,10] in the application studied as well as the inference scheme. Both [9,10] formulate the localization as a Bayesian inference problem and approximate the posterior distributions of interest with Gaussians. [10] uses a moment matching method and appears to be centralized in nature.…”

“…[10] uses a moment matching method and appears to be centralized in nature. The method in [9] uses instead linearization, is distributed and on-line, but its implementation relies on communication via a junction tree (see [13] for details) and requires an anchor node as pointed out in [14,Section 6.2.3]. In this paper we formulate the sensor localization problem as a static parameter estimation problem for Hidden Markov Models (HMMs) [15,16] and we estimate these static parameters using a ML approach, which has not been previously developed for the self-localization problem.…”

“…A clear advantage of this approach compared to previous alternatives is that it makes an on-line implementation feasible. Finally, [11] is based on the principle shared by our approach and [9,10]. In [11] the authors exploit the correlation of the measurements made by the various sensors of a hidden spatial process to perform self-localization.…”

“…The value of a global static parameter is measured at each node via a linear Gaussian observation model and the aim is to obtain a maximum likelihood estimate in a distributed fashion. Note that all the aforementioned papers, except [9] and [10], do not deal with a distributed localization and tracking task.…”